ACM Presents: Goodput as a Function of Operational Maturity: A Process-Engineering Framework for AI Infrastructure

About this event

Goodput is the fraction of time that accelerated compute spends doing useful training or inference. It has become the main measure of how productive AI infrastructure is. Industry analysts predict that about 80% of current neocloud providers will stop operating within two years. That raises questions about how long both providers and their enterprise customers can keep running. This talk will make the case that high-goodput AI infrastructure operators treat operations as an engineering discipline rather than a craft. Leading operators also insert an abstraction layer for managing bare metal between the hardware layer and the platform layer.

RSVP: https://pin.gsu.edu/event/12875322

About the Speaker: Tom Gillman is an IT professional with over 40 years of experience in all aspects of data and compute infrastructure. He has worked for ATT, VMware, Dell, Microsoft, EMC, and other companies doing large-scale machine and data center design, implementation, and operations. His focus for the last several years has been in developing platform automation at scale, specifically focusing on building machine clusters with emphasis on AI inference and training. Tom is a 2002 GSU graduate with dual degrees in Computer Science and Mathematics and has lived in the Atlanta area for most of his professional career. Tom’s personal hobbies include reading, computer gaming, and martial arts.

Everyone welcome!

Free refreshments!